How the Demand for Servers Is Changing Amid the Rise of AI

Servers are the backbone of the IT infrastructure of many enterprises, including cloud providers that power countless businesses. Server demand is experiencing a shift, with the total number of shipments in 2023 expected to decline from previous years, according to a report from London-based Omdia, a research and consultancy company.

More importantly, the kind of server enterprises are going for is also changing. With artificial intelligence (AI) rapidly being adopted by companies in various industries, the demand for servers is driven by demand for AI. In simple language, server procurement is changing as enterprises look to invest in AI-friendly equipment. 

In this article, we will discuss the following:

• The findings from the Omdia report.
• Details on server sales.
• How AI impacts the demand for servers.

Not the article you were looking for today? Try these out:

• The Importance of Maintenance Audits ⁠— And How to Carry Them Out
• How to Make the Best Choice on White Box Servers for Your Data Center
• Buying a Server for Your Enterprise: 8 Things to Know

Server Shipments to Slow Down

Omdia reported a decline in the sales of servers beginning with the first quarter (Q1) of 2023. In Q1, 2.8 million servers were sold, below the expected 3.2 million. The research firm reports that the overall sales throughout the year will continue to decline. 

This is unprecedented, considering that server sales have been consistently rising since 2007. That was the last time the sales numbers dropped, thanks to the recession. 

The report listed several possible reasons behind the reduction in server shipments. A highly likely reason is the uncertain economic climate. Another reason is the overall budget cuts in IT spending.  

In January, Gartner had already cut its IT spending growth estimates by half to just 2.4 percent, or $4.5 trillion. Before the last quarter of 2022, its prediction was 5.1 percent growth. 

Omdia also suggested that server sales could be down because companies delay refresh cycles. The type of servers they now want to invest in can cost upwards of half a million dollars per unit. 

The company also suggests that companies are acting cautiously at the moment, relying on cloud services for now and waiting to assess their business demands for the future before making any significant infrastructure investments. 

Servers With GPUs in More Demand

Interestingly, the demand for more advanced, more expensive servers is high. 

The report suggests that interest in servers with GPUs is more in demand than regular servers with just CPUs. This is mainly because of the increasing interest in developing and hosting AI applications. 

During Q1 of 2023, Microsoft, Google, and Meta procured AI-ready servers. These tech giants quickly invested in developing AI chatbots to rival ChatGPT, OpenAI’s successful chatbot, relying on the company’s most advanced generative AI model, GPT 4. One can put two and two together to see why companies like Microsoft and Google purchased new servers at the beginning of the year. 

The demand for compute resources is strong and is expected to remain strong through the end of the year and beyond. While overall server sales may be low, increasing demand for computing means that servers with GPUs will quickly dominate the market share. 

The race to adopt and use AI by significant tech corporations, cloud providers, and startups is leading to the rise in the sale of these expensive servers. 

The ROI on AI-Related Hardware Investments

In another report, Omdia said AI investments produce quick returns on investment (ROI). Its survey asked respondents whether AI had increased topline revenue, reduced the bottom line, or improved efficiency and customer experience. The findings confirmed the potential of AI to boost revenues. 

Nearly 54 percent of respondents reported an ROI of one percent, while 14 percent reported an ROI of 11 percent or more. 

This survey was conducted in early 2023 when most enterprises had just begun adopting AI. The takeaway from the survey is that AI investments aren’t just returning profit but doing so quickly. The ROI percentage would likely improve even further when the companies have had the technology for longer. 

Omdia expects AI spending to rise as more companies rush to adopt it and tap the innovative and economic benefits of the technology. It expects the ROI to rise for companies working with AI for over a year. 

And as companies see success with one AI project, they’re likely to expand their deployments into other areas. 

This report’s findings align with the actual sales data of servers, showing an increased demand for compute-heavy units with multiple GPUs. 

Is It a Good Time to Upgrade Your Servers?

For many enterprises and service providers with server refresh cycles on the horizon, the question is whether to buy the typical enterprise servers or invest in powerful but expensive AI-ready units. Even if it’s not yet the time to replace servers, the race for AI adoption and supremacy may entice companies to refresh earlier. 

So what should you do if you find your enterprise in such a situation? Should you replace existing servers with more advanced ones with more computing power to perform tasks like AI model training?

Here are three factors to consider when making this decision:

Lifespan of Current Servers

Assess the current state of your servers and estimate the remaining lifespan of your servers. The Omdia report on server sales also revealed that the average useful life of a server is 6.4 years now, up from previous years. In other words, many companies are keeping their servers for longer (or they’re lasting longer). 

Many enterprises deploy servers beyond the end of service life as per the manufacturer, with support from third-party maintenance (TPM) providers. 

If your servers have a considerable life span left, it may make sense to wait out and purchase more advanced servers later. 

On the other hand, one can sell these servers to buy compute-rich servers. Remember that you’d still need a significant investment as the latter is much more expensive. 

Business Needs

Your business requirements and strategy should be the primary factor for this decision. While AI is beneficial in various ways, it may only be necessary for some IT enterprises. 

Additionally, many businesses are dabbling with AI as a service to try out technologies like machine learning before investing in expensive infrastructure to have on-premise. 

Take the time to assess how AI can transform your business operations or the service you provide to your customers before you decide to purchase AI-friendly servers. 

CapEx Budget

Lastly, your capital expenditure (CapEx) budget should also help you make the call for server refresh. If you have access to capital for implementing significant infrastructure changes, investing in servers with GPU may make sense. On the other hand, if you have a restricted CapEx budget, you may have little choice but to stick with the servers you have for the foreseeable future. 

The Takeaway

Omdia’s findings show that overall server sales will decrease in 2023. However, the sales of servers with more computing resources, i.e., GPUs, are poised to rise on the back of the AI revolution. 

The fear of missing out on AI adoption fuels many enterprises, especially big tech, to buy advanced, expensive servers. And early investments are showing high ROIs. That all said, buying new servers is a significant investment that must be made with ample planning and future assessment. 

PivIT can help you decide whether procuring new servers suits your enterprise. Consult with a procurement specialist today!